Quantitative ultrasound in relation to risk factors for low bone mineral density in South African pre-menopausal women

The study nurses approached women in the centers to determine if they met eligibility requirements for the study and if they wanted to participate. The nurses recruited the women without knowledge of their contraceptive use histories.

Women eligible for the study were aged 18–44, pre-menopausal, had not been pregnant or had not breastfed in the previous year, and did not have an illness or use medication that would influence their bone health. Medical conditions for exclusion were illnesses requiring bed rest for more than 6 weeks in the past 12 months, thyroid, parathyroid, and pituitary disease, cancer, sarcoidosis, rheumatoid arthritis, chronic liver and kidney disease, rickets, Paget’s disease, and osteoporosis. Women taking anticonvulsants, systemic corticosteroids, drugs for hypo-, hyper-, or parathyroidism, thiazide diuretics, or calcium supplements for more than 6 months in the last 5 years were also excluded. Written informed consent was obtained from all subjects.

Of the 4,362 women approached, 3,957 were willing to participate in the study. Four hundred and thirteen were additionally excluded for failing to meet study criteria or because the Sahara measurement failed, leaving 3,544 women. We excluded 51 from the analysis because of ultrasound measures below or above the 0.5 percentile of measures (broadband ultrasonic attenuation [BUA] ≤45 and ≥117 dB/MHz or speed of sound (SOS) ≤1,500 and ≥1,645.5 m/s), leaving 3,493 women; 1,598 were black and 1,895 were of mixed racial ancestry, which could be a combination of any of the following: black, white, Asian, or Khoisan ancestry.

The study nurses administered standard questionnaires to collect information on demographic, reproductive, and contraceptive history, and smoking and alcohol intake. We categorized alcohol intake and smoking in terms of current, past, or never use. For both smokers and drinkers, current and past users were asked information on duration, frequency, and quantity of intake. Past drinkers and smokers were also asked number of years since last use. Current intake of selected calcium-rich foods was obtained using a questionnaire that has been validated against 3-day records (Micklesfield, unpublished data). Information on the frequency and quantity of milk, yogurt, cheese, and fish (such as pilchards which are rich in calcium) intake was obtained and the total number of weekly servings for these items was calculated. The physical activity section of the questionnaire was adapted from Kriska et al. [28] and has previously been applied within the South African context [29]. Historical information on physical activity was obtained for three epochs—primary school, high school, and post-school. The nurses recorded walking to school and other habitual walking activities lasting longer than 40 min, including walking for transport, for pleasure, and walking herding cattle, carrying wood, or fetching water. They also recorded participation in sports and leisure activities such as athletics, tennis, netball, volleyball, and dancing. In addition, for participants who had worked in jobs for 2 years or more, information was collected on walking and carrying heavy objects at work.

To take account of load bearing during activities, the impact of loading from each physical activity was ranked on a 0–3 scale with “0” indicating non-weight-bearing (e.g., swimming) and “3” indicating high impact such as playing volleyball [30]. Total impact hours (TIH) were calculated for each of the three epochs.

Weight and height measurements were taken using a SECA (Hamburg, Germany) standard floor scale and SECA height measure with participants wearing light clothing and no shoes. Body mass index (BMI) was calculated as weight divided by height squared in kilograms per square meter.

Bone mass was assessed using a heel gel-coupled (dry) quantitative ultrasound system, the Sahara (Hologic, SN 03281, SN 03278). The Sahara ultrasound device measures two parameters: speed of sound (SOS) in meters per second and broadband ultrasonic attenuation (BUA) in decibels per megahertz. SOS is the distance between the two transducers divided by the time it takes for the signal to pass from one transducer through the heel to the opposite transducer. BUA is the slope of the regression line for the relationship between the ultrasound attenuation and the sound frequency over the range 0.1–1 MHz [31]. Higher values of BUA and SOS are associated with greater bone mass. The quantitative ultrasound index (QUI) is a linear combination of both BUA and SOS \( \left( {{\text{QUI}} = \left( {{\text{BUA}} + {\text{SOS}}} \right) \times 0.{\text{41}}-{\text{571}}} \right) \). BUA is correlated with heel BMD as measured by DXA (r values ~0.8) [32].

The study nurses were trained in using the QUS machines until they achieved proficiency such that the coefficient of variation (CV) for repeated measures of the same subject was within the machine specification as specified by Hologic. For the Mitchells Plain machine, the CV for 40 repeated measures was 2.7% for BUA and 0.3% for SOS. For the Gugulethu machine, the CV was 2.1% for BUA and 0.3% for SOS for 21 repeated measures. On a daily basis, a phantom was used to check the quality control of each machine. The phantom values were required to fall within the quality control (QC) limits of 0.86–1.14 for QAB and 0.986–1.014 for QAS. QAB and QAS are dimensionless quantities calculated by dividing the phantom BUA and SOS measured in the daily quality control procedure by the values specified for that phantom by the manufacturer. If the machine did not pass the QC, it was recalibrated. If it continued to fail QC, the transducer pads were replaced. In addition, if, over a number of weeks, the mean QC values for a machine drifted outside Sahara specification, the machine was also recalibrated. Both machines were recalibrated twice over the study period.

To assess correlation of the QUS measures with DXA measures, a sample of 14 women was measured on both the Gugulethu and the Mitchells Plain QUS machines and in addition had a DXA scan (DXA Hologic, Model Discovery W S/N 80196) performed at the same visit. For the Gugulethu sonometer, the correlation coefficients between BMD assessed by DXA at the femoral neck, femoral trochanter, total hip (the combination of femoral neck, femoral trochanter, and intertrochanteric areas), and total lumbar spine (L1–4) and BUA of the calcaneus were 0.69, 0.78, 0.74, and 0.60; the corresponding correlation coefficients between BMD and SOS were 0.65, 0.84, 0.73, and 0.54. For QUI, these coefficients were 0.68, 0.84, 0.75, and 0.58 (all significant; p < 0.05). Similarly, for the Mitchell’s Plain sonometer, the correlation coefficients of the four DXA measurements with BUA were: 0.72, 0.77, 0.74, and 0.53; with SOS they were 0.62, 0.82, 0.70, and 0.50; and with QUI they were 0.67, 0.83, 0.74, and 0.52 (all p < 0.075).

We calculated the crude mean value of BUA, SOS, and QUI for each category of a specific risk factor and obtained adjusted mean of BUA, SOS, and QUI using linear regression models. Variables included in the multivariable regression model were age, height, BMI, education, age at menarche, number of live births, smoking and alcohol use, calcium intake, and early physical activity. Adding contraceptive use to the model did not affect the results. We tested the linear dose–response relationship by entering the median value for the exposure category of interest into a term in the regression model. To further investigate residual confounding by BMI of height, physical activity, and age at menarche, the multivariable regression model was again tested according to three strata of BMI. All analyses were performed using STATA software version 9.0 (StataCorp LP, TX, USA)